Safety Valves vs Relief Valves: Sizing, Certification, and Venting Safely

Every pressurised system needs a last line of defence: a self-actuating valve that opens when the pressure exceeds what the equipment can safely contain, and that does so with no power, no instrumentation, and no operator. That device goes by several names — safety valve, relief valve, safety relief valve, PSV, PRV — and the names are not interchangeable. They describe genuinely different opening behaviours, suited to different services, governed by different code sections.

This guide covers the terminology, where each type belongs, how sizing works under API 520/521, the certification stamps to look for, and the installation rules that determine whether a correctly sized valve actually protects anything.

One caution before the details: pressure relief is safety-critical, code-governed work. This article explains the concepts and the standard methods, but final valve selection must be based on the manufacturer's certified capacity data and verified against the code that applies to your equipment, by the engineer responsible for it.

Terminology: relief valve, safety valve, safety relief valve

API 520 and the ASME Boiler and Pressure Vessel Code distinguish three devices, all spring-loaded, all opening on inlet static pressure:

• Relief valve — opens in proportion to the overpressure. As pressure rises above the set point, the disc lifts progressively, and capacity increases gradually. Intended primarily for incompressible (liquid) service, where relieving even a small volume drops the pressure quickly.
• Safety valve — characterised by rapid full opening, or "pop" action. A huddling chamber uses the escaping vapour to snap the disc to full lift within a few percent of set pressure. Intended for compressible service — gas, vapour, and above all steam — where a partially open valve would relieve too slowly to arrest a pressure rise.
• Safety relief valve — built to behave either way: pop action on gas, proportional opening on liquid. This is the workhorse of process plants, since a vessel's relief case may involve vapour, liquid, or both.

Two abbreviations cause endless confusion:

• PSV (pressure safety valve) is the common plant tag prefix for all of the above.
• PRV is used by API as the umbrella term "pressure relief valve" — but in piping and building-services contexts PRV very often means a pressure reducing valve, a regulator that throttles continuously to hold a downstream set pressure during normal operation. A reducing valve is a control device; a relief valve is a protection device. One works all day; the other should almost never lift. Never let one substitute for the other, and when reading a P&ID from another discipline, check which PRV is meant.

In European practice under EN ISO 4126, "safety valve" is the umbrella term for the whole family, which is worth knowing when reading datasheets from EU manufacturers.

Beyond the basic spring-loaded types there are two important construction variants:

• Balanced bellows valves isolate the spring from the discharge side with a bellows, so back pressure in the outlet piping does not add to the spring force.
• Pilot-operated valves use process pressure, routed through a small pilot, to hold the main disc closed — giving very tight seating close to set pressure and high tolerance to back pressure.

Why that matters comes up in the venting section below.

Where each type belongs

• Steam boilers and steam drums — safety valves (pop action), governed by ASME Section I. Speed of opening is the whole point: boiler heat input does not stop when the valve starts to lift.
• Air receivers, gas storage, process vessels — safety or safety relief valves under ASME Section VIII (or PED in Europe). Compressed-air receivers are the most common relief application most engineers meet.
• Liquid-full equipment and blocked-in piping — relief valves. The classic case is thermal relief: a liquid-packed line blocked in between two valves and warmed by the sun can build enormous pressure from a tiny expansion volume, so a small proportional-opening valve is enough.
• Positive-displacement pump and compressor discharge — a relief valve (or internal bypass) is mandatory, because a PD machine against a closed discharge will keep raising pressure until something yields.
• Alongside rupture discs — discs can stand alone, or sit upstream of a relief valve to isolate it from corrosive or fouling fluid (this combination carries a 0.9 capacity penalty in sizing, the factor $K_c$ below).

The vocabulary of overpressure

Sizing conversations are conducted in a handful of precisely defined terms:

• MAWP — maximum allowable working pressure, the pressure rating of the protected equipment. The anchor for everything else.
• Set pressure — where the valve begins to open. For a single valve this must not exceed the MAWP.
• Overpressure — the pressure rise above set pressure at which the valve achieves rated capacity, expressed as a percentage of set pressure.
• Accumulation — the pressure rise above MAWP during relief. ASME Section VIII permits 10% accumulation for a single valve in a non-fire case, 16% for multiple valves, and 21% for the external-fire case. Section I boilers are held tighter, at 6%.
• Blowdown — how far pressure must fall below set pressure before the valve reseats, typically several percent. Too little blowdown and the valve cycles; too much and it stays open wastefully long.
• Back pressure — pressure at the valve outlet, with two components: superimposed back pressure that exists before the valve opens (a pressurised relief header, for instance) and built-up back pressure created by the valve's own discharge flowing through the outlet piping.

Sizing under API 520 and API 521

Relief valve sizing is a four-step process, and the valve equation is the last step, not the first.

Step 1 — identify the overpressure scenarios

API 521 catalogues the credible causes of overpressure: blocked outlet, external fire, control valve failing open, cooling or reflux failure, heat exchanger tube rupture, thermal expansion, utility failure, and more. Each scenario implies a different relieving rate. The governing case is whichever scenario demands the largest orifice — not necessarily the one with the largest flow, since fire cases relieve at higher accumulation than operational cases.

Step 2 — establish the required relieving rate

For a blocked outlet, the relieving rate is the maximum the upstream system can deliver. For fire, API 521 gives wetted-area heat-input correlations from which a vapour generation rate follows. For thermal relief, it follows from the heating rate and the liquid's expansion coefficient. This step is where most of the real engineering lives.

Step 3 — compute the required orifice area

For gas or vapour in critical (choked) flow — the usual case, since choking occurs whenever the outlet pressure is below roughly half the relieving pressure — API 520 gives, in SI units:

• $A$ — required orifice area (mm²)
• $W$ — required relieving rate (kg/h)
• $P_1$ — relieving pressure, kPa absolute (set pressure + allowed overpressure + atmospheric)
• $T$ — relieving temperature (K)
• $Z$ — compressibility factor at relieving conditions
• $M$ — molecular mass of the gas

The coefficient $C$ depends only on the ideal-gas specific-heat ratio $k$:

which gives $C \approx 0.027$ for air ($k = 1.4$). Flow is critical when the back pressure is below

— about 0.53 × $P_1$ for air. The correction factors:

• $K_d$ — effective discharge coefficient (0.975 for gas service in preliminary sizing)
• $K_b$ — back-pressure correction factor
• $K_c$ — 0.9 when a rupture disc is installed upstream, 1.0 otherwise

For liquid service, the API 520 form is:

• $Q$ — required relieving rate (L/min)
• $P_1 - P_2$ — relieving pressure minus total back pressure (kPa)
• $G$ — specific gravity relative to water
• $K_d$ — effective discharge coefficient, 0.65 for liquid trim
• $K_w$ — back-pressure correction for balanced valves
• $K_v$ — viscosity correction, significant for viscous fluids

Steam has its own equation built on Napier's formula, with corrections for superheat and for high set pressures — conceptually identical: required area from mass rate and relieving pressure.

Two-phase relief — flashing liquids, reactive cases — is a specialist topic. API 520 Annex C and the DIERS methodology cover it; none of the single-phase equations above are safe to extrapolate there.

Step 4 — select a standard orifice

You do not buy an orifice of exactly the computed area. API 526 standardises relief valve orifices in lettered sizes from D (71 mm²) to T (16,774 mm²), with standardised centre-to-face dimensions and flange ratings. You select the next letter larger than the required area. A subtlety worth knowing: the API "effective" areas and effective $K_d$ are for preliminary sizing; the final check must use the manufacturer's actual orifice area and certified discharge coefficient together, never a mix of the two systems.

Resist the temptation to add margin by jumping extra letters. An oversized relief valve is not conservative — it lifts, dumps far more than the system is supplying, collapses the inlet pressure, slams shut, and repeats. That cycle is chatter, and it destroys seats and can fatigue the inlet nozzle.

All three service equations — with the orifice-letter selection and the back-pressure, superheat, and viscosity corrections — are implemented in our free relief valve sizing calculator, which is regression-tested against the standard's own worked examples.

Certifications: what the stamps mean

Relief devices are one of the few piping components where certification is not paperwork — the capacity number on the nameplate is a tested, audited value.

• ASME Section I ("V" stamp) — safety valves for fired boilers. Tightest accumulation limits, pop action mandatory.
• ASME Section VIII ("UV" stamp) — relief devices for unfired pressure vessels. The stamp certifies that the valve's rated capacity is achieved at 10% overpressure.
• National Board certification — in the ASME system, capacities are certified through National Board testing, and the rated capacity is set at 90% of the average tested flow — a deliberate, built-in conservatism. Certified valves and their capacities are listed in the National Board's public "NB-18" listing.
• API 526 / API 527 — API 526 standardises dimensions and orifice letters so valves are interchangeable across manufacturers; API 527 defines the seat tightness test (how little a closed valve may leak near set pressure).
• PED 2014/68/EU (Europe) — relief devices are "safety accessories" and fall into Category IV, the most stringent conformity route, with notified-body involvement and CE marking. The harmonised design standard is EN ISO 4126. A valve sold for protection duty in the EU must carry the CE mark under PED; an ASME stamp alone does not satisfy the directive, and vice versa — equipment traded across both markets often carries dual certification.

For the buyer, the practical rule is simple: the nameplate must show a recognised certification mark, the set pressure, and the certified capacity at the stated overpressure, and the sizing file must reference that certified capacity — not a catalogue estimate.

Installation and venting safely

A perfectly sized valve is routinely wrecked by the piping around it. Two numbers govern the installation, and both are ordinary pressure-drop calculations.

The 3% inlet rule

API 520 Part II limits the non-recoverable pressure loss in the inlet line to 3% of set pressure at rated flow. The reason is dynamic: when the valve pops, full flow suddenly runs through the inlet stub, the pressure at the valve nozzle drops by the line loss, and if that drop eats the blowdown margin the valve reseats — then pressure recovers and it pops again. That is chatter driven not by oversizing but by inlet piping. Keep the inlet short and direct, at least the valve inlet size with no reductions, and count the entrance loss from the vessel nozzle.

Back pressure and the 10% rule

On the outlet side, the discharge flow generates built-up back pressure in the tailpipe or relief header. For a conventional spring valve, back pressure acts on the disc and adds to the closing force, so the variable (built-up) component is limited to 10% of set pressure — beyond that, capacity falls and stability suffers. When the total back pressure must be higher — long headers, flare systems, or superimposed pressure that varies — select a balanced bellows valve (typically comfortable to 30–50% back pressure) or a pilot-operated valve (higher still). This is a selection decision made from a piping calculation: you cannot pick the right valve construction until you know the back pressure its own discharge will create.

Discharging to a safe place

Where the relieved fluid goes is as important as whether the valve opens:

• Steam and air can vent to atmosphere through a vertical stack discharging above roof level and away from platforms — positioned so the jet, the noise, and (for steam) the condensate cannot reach people.
• Flammable or toxic gases go to a closed relief header and on to a flare, scrubber, or knock-out drum — sized as a network, because several valves may relieve simultaneously and each one's back pressure depends on the others.
• Every open vent needs a drain hole or drip leg at the low point: rainwater standing on top of the valve disc corrodes the seat and can freeze. Loose-fitting rain caps that cannot restrict flow are acceptable; anything that could stick shut is not.
• Reaction force: a large valve at full flow produces a jet thrust at the first elbow that can bend inadequately supported tailpipes. API 520 gives the force estimate; brace the discharge piping for it.
• Never value-engineer away access. Valves need periodic testing (set pressure verification and reseat) — typically on a fixed inspection interval — and that requires reaching them.

The isolation valve question

Anything that can isolate a relief valve from the equipment it protects defeats it. Where block valves are unavoidable — to allow online valve maintenance with a spare — they must be full bore and locked or car-sealed open, under a documented administrative procedure, with an interlock ensuring one relief path is always in service. Many jurisdictions regulate this explicitly; treat any valve between a vessel and its PSV as a design decision needing justification, not a convenience.

Common mistakes, quickly

• Sizing for the largest flow scenario instead of the largest area scenario.
• Adding "safety margin" by oversizing the orifice → chatter.
• Ignoring inlet line loss (breaks the 3% rule) → chatter.
• Selecting a conventional valve, then routing it into a header whose built-up back pressure exceeds 10% of set → capacity loss and instability.
• Mixing API effective areas with manufacturer certified coefficients.
• Forgetting $K_c = 0.9$ when a rupture disc sits upstream.
• Treating a pressure reducing valve as protection, or assuming a control valve "would never fail open".
• Extrapolating single-phase equations to flashing (two-phase) relief.

Where a network simulator fits in

To be clear about scope: SimuPipe does not size relief valves — sizing is done against certified capacity data under the codes above. But both of the numbers that decide whether an installation works are pipe-flow problems, and they are exactly the kind a network simulator answers:

• the inlet loss at rated flow for the 3% rule — a short compressible-flow calculation where the entrance loss and any fittings dominate, and
• the built-up back pressure at the valve outlet — a tailpipe or header pressure-drop calculation at relieving conditions, which decides between a conventional and a balanced valve.

Size the orifice itself in the relief valve sizing calculator, check a tailpipe in the free pressure drop calculator, size the vent line bore in the pipe sizing calculator, or model the whole discharge header — several sources relieving into one line, with real gas properties and choking checked — in the simulator sandbox, no account needed. For the valves that work every day rather than the ones that should never lift, our guide to control valve sizing under IEC 60534 covers the other side of the family.